The Structure of Misfolded Amyloidogenic Dimers: Computational Analysis of Force Spectroscopy Data

Abstract

Progress in understanding the molecular mechanism of self-assembly of amyloidogenic proteins and peptides requires knowledge about their structure in misfolded states. Structural studies of amyloid aggregates formed during the early aggregation stage are very limited. Atomic force microscopy (AFM) spectroscopy is widely used to analyze misfolded proteins and peptides, but the structural characterization of transiently formed misfolded dimers is limited by the lack of computational approaches that allow direct comparison with AFM experiments. Steered molecular dynamics (SMD) simulation is capable of modeling force spectroscopy experiments, but the modeling requires pulling rates 10⁷ times higher than those used in AFM experiments. In this study, we describe a computational all-atom Monte Carlo pulling (MCP) approach that enables us to model results at pulling rates comparable to those used in AFM pulling experiments. We tested the approach by modeling pulling experimental data for I91 from titin I-band (PDB ID: 1TIT) and ubiquitin (PDB ID: 1UBQ). We then used MCP to analyze AFM spectroscopy experiments that probed the interaction of the peptides [Q6C] Sup35 (6–13) and [H13C] Aβ (13–23). A comparison of experimental results with the computational data for the Sup35 dimer with out-of-register and in-register arrangements of β-sheets suggests that Sup35 monomers adopt an out-of-register arrangement in the dimer. A similar analysis performed for Aβ peptide demonstrates that the out-of-register antiparallel β-sheet arrangement of monomers also occurs in this peptide. Although the rupture of hydrogen bonds is the major contributor to dimer dissociation, the aromatic-aromatic interaction also contributes to the dimer rupture process.

Introduction

Self-assembly of proteins and peptides into nano-assemblies is a ubiquitous phenomenon that is of great importance to biology. It is assumed that such a process is accompanied by the transition of proteins into misfolded states, but knowledge about this process is very limited. Traditional structural analytical methods are not amenable to probing transient misfolded states of proteins, but progress has been made with the use of single-molecule biophysics methods (reviewed in Lyubchenko et al. (1)). Single-molecule force spectroscopy (SMFS) combined with atomic force microscopy (AFM) has been a powerful tool for probing protein mechanics and characterizing protein folds (2). The use of the SMFS approach, in which the interaction of two identical protein systems is probed, made it possible to characterize the interactions of amyloid proteins and misfolded peptides (3, 4, 5, 6, 7). Such studies revealed a number of important properties of amyloid proteins in transient dimeric states. Dynamic force spectroscopy analysis led to the conclusion that misfolded states of proteins are stabilized dramatically when they assemble into dimers. The lifetimes of dimers are on a timescale of seconds and thus are many orders of magnitude higher than the lifetimes of transient protein states. Importantly, the high stability of dimers does not significantly depend on the protein or peptide size; therefore, the elevated stability of misfolded dimers was observed for peptides as small as the Sup35 heptamer (6), amyloid β protein 40 and 42 (5, 7), and α-synuclein protein (140 residues (3, 4)). The contour length derived from AFM force spectroscopy experiments provides information about the location of interacting segments within the dimers (3, 4, 8, 9, 10, 11); however, the structural characterization of the dimers remains unclear. Comparing AFM results with a theory capable of describing the experimental phenomena would be one way to extract the necessary structural information. In steered molecular dynamics (SMD) simulations, an external force is applied to classic molecular dynamics (MD) simulations that allow AFM pulling experiments to be modeled (12, 13, 14, 15, 16). However, a major problem with this approach is that SMD is typically performed at pulling rates of 5 nm/ns, which are 10⁷ times higher than those used in typical experimental conditions (14, 17), thereby complicating a direct comparison between in silico data and the experimental results (18). In a recent publication, high-speed AFM instrumentation used in experimental conditions approached the SMD temporal range (19), but it is not in common use at this time. Recently, Jónsson et al. (20) described an all-atom Monte Carlo pulling (MCP) approach that enabled modeling at rates comparable to those obtained in regular AFM pulling experiments (pulling rate ∼300 nm/s). The authors were able to directly compare their results with AFM experiments for amyloid β and α-synuclein monomers.

In this study, we further developed the MCP approach, which enabled us to model AFM pulling experiments that measured interpeptide and intrapeptide interactions. MCP analysis of AFM probing experiments for peptides Sup35 (6–13) ([Q6C] Sup35 (6–13); sequence: CGNNQQNY) and Aβ (13–23) ([H13C] Aβ (13–23); sequence: CHQKLVFFAED) demonstrates that both peptides in the AFM experiments form transient misfolded dimers with an antiparallel orientation of the monomers’ β-sheets. The monomers are arranged in an out-of-register pattern with an overall length of interacting segments of five residues. The MCP approach also enabled us to follow the rupture process and characterize the contribution of different interactions to dimer stability.

Materials and Methods

Initial structures

Two peptides, CHQKLVFFAED (Aβ peptide) and CGNNQQNY (Sup35 peptide), were studied. The structure of the Aβ dimer was taken from our previous publication (14) and is shown in Fig. 1 A. The Aβ dimer with an in-register β-sheet in Fig. 1 B was generated by the Protein Folding and Aggregation Simulator (PROFASI) software package (21). The structures of the Sup35 dimers (the out-of-register β-sheet and the in-register β-sheet) were obtained from replica exchange MD (REMD) (22) simulations and are shown in Fig. 1, C and D, and Fig. S1 in the Supporting Material. The REMD simulations were performed by using the GROMACS 4.5.5 package (23) with the AMBER-ff99SB-ILDN force field (24). Simulation parameters were adopted, with modifications, from our previous publication (14). The complete simulation details can be found in the Supporting Material.

Figure 1.

Initial structures for the MCP simulations. (A) The out-of-register Aβ dimer structure obtained in our previous study (14), with an antiparallel orientation of the monomers, was chosen for the MCP simulation. (B) The in-register Aβ dimer generated in this work by using PROFASI software. (C and D) The antiparallel out-of-register (C) and in-register (D) Sup35 dimers correspond to structures with the lowest-energy minima in the REMD simulations. The stick structures correspond to the backbones of the two monomers, and the dotted lines represent H-bonds. The balls indicate the Cα atoms for the N-terminal residues of the monomers, where the pulling force was applied. To see this figure in color, go online.

MCP simulation

We implemented the modified MCP simulations using the PROFASI package (21) with the implicit water all-atom model and the FF08 force field. The modification details are provided in the Supporting Material.

The total energy during the pulling process was calculated according to the following equation:

$${\text{E}}_{\text{tot}}=\text{E}\left(\text{x}\right)+\frac{\text{k}}{2}{\left[{\text{L}}_0+\text{vt}-\text{L}\left(\text{x}\right)\right]}^2,$$

where E(x) is the energy in the absence of an external force, t is MC time, and k is the spring constant of the probe. L₀ represents the distance between the Cα atoms of Cys residues at the N-termini from the initial conformation. L(x) is the distance between Cα atoms of Cys residues during MC pulling, and x denotes a protein conformation. When v = 0.1 fm per MC step, the value is equivalent to 600 nm/s. The parameters for each case are listed in Table S1.

Data analysis

Several hundred simulations were performed and the results were assembled in Table S1. The force curves were smoothed by MATLAB 2013a (The MathWorks, Natick, MA). The rupture peak was defined by a force value > 20 pN, and the position of the peak was identified by the minimum derivative value of the smoothed force curve. Rupture force distributions for each structure were compared with experimental values and fitted by probability density function (PDF) (25, 26). The majority of simulations were fitted by bimodal PDF, indicating the existence of the transient states of the dimers, but the results from the in-register dimers at low temperatures were fitted by unimodal PDF. The Kolmogorov-Smirnov nonparametric test (SPSS 20.0; IBM, Armonk, NY) was used to determine the statistical significance of the differences in force distribution. The fractions of dissociated dimers for both peptides, for in-register and out-of-register conformations, were obtained by dividing the number of simulations with nonrupture events by the total number of simulations.

Results and Discussion

MCP approach

The all-atom MCP method was previously described to unravel the intramolecular structure of proteins by pulling apart the N- and C-terminal residues (20). We modified this approach for AFM probing experiments in which the dimer, formed by two monomers immobilized to the AFM tip and the substrate, is pulled apart. With these modifications, we are able to apply pulling forces to any pair of Cα atoms. We tested the approach by using experimental data for two commonly used experimental systems: titin I91 (formerly I27) and ubiquitin proteins. Each repeated unit of the I91 protein unravels in a stepwise pattern under the applied force. In the experiment, with a pulling rate of 600 nm/s (2, 27), each segment of I91 protein ruptures cooperatively, producing a rupture force value of 200 ± 26 pN. Similar experimental studies for ubiquitin (28) resulted in a rupture force value of 203 ± 35 pN.

We used our MCP approach to model the pulling process of one unit of I91 by using the available PDB structure (PDB ID: 1TIT (29)). A typical force curve for the rupture of this I91 unit is shown in Fig. 2 A. The mean rupture force values are 184 ± 37 pN (n = 140) at a pulling rate of 600 nm/s, and 203 ± 33 pN (n = 162) at a pulling rate of 1 μm/s. The experimental value of 200 ± 26 pN obtained at a pulling rate of 600 nm/s (2, 27) is very close to both theoretical values. Recent in silico results obtained with the coarse-grained model at a pulling rate of 600 nm/s produced a rupture force value of 204 ± 30 pN (30), which is also close to our results. Similarly, we modeled ubiquitin rupture by using the PDB structure (PDB ID: 1UBQ (31)). The results obtained at a pulling rate of 400 nm/s are shown in Fig. 2 B. The maximum rupture force of 208 ± 51 pN (n = 199) is very close to the experimental value reported by Carrion-Vazquez et al. (28) (203 ± 35 pN) and the in silico value obtained by Sikora et al. (30) (230 ± 34 pN). Therefore, our MCP approach produces pulling results that are in agreement with experimental data.

Figure 2.

Typical force curves for unraveling I91 domains and ubiquitin. (A) The initial structure of I91 was taken from the PDB website (PDB ID: 1TIT). The snapshot just before rupture is on the right. The rupture force is 200 pN, followed by the breakage of H-bonds within the β-strands of A′-G, as shown on the right. (B) Unfolding of ubiquitin (PDB ID: 1UBQ). The snapshot before the maximum rupture is on the right. The rupture event occurs at the breakage of H-bonds between β-strands I and V, as shown on the right side of the force curve. In the schematics of the structures, the arrows indicate β-strands, the tubes are random coils of different types, and the ribbon represents α-helix. To see this figure in color, go online.

Structural features of Aβ and Sup35 peptides

We used our MCP approach to analyze the dimers formed by two amyloidogenic peptides, Aβ and Sup35. Both peptides were probed in SMFS experiments in which each monomer was tethered to the AFM tip and substrate surface via terminal Cys residues. The dimers’ dissociation was characterized by sharp rupture events with forces in the range of 100 pN (6, 14). Therefore, to closely mimic the AFM experimental conditions during the MCP simulations, Cys residues were added to the N-termini of the peptides and the pulling force was applied at these points.

In the computational analysis, we chose four different conformers (Fig. 1) as the initial structures for the MCP analyses. For the Aβ dimer, we selected the structure generated from our previous publication (14). In this structure, shown in Fig. 1 A, the two monomers adopt an out-of-register antiparallel β-sheet conformation stabilized by four backbone hydrogen bonds (H-bonds). H-bonds are formed between residues His-14 of monomer A and Phe-19 of monomer B, and between Lys-16 of monomer A and Leu-17 of monomer B. Salt bridges and aromatic interactions are also involved in stabilizing the Aβ dimer structure (14). Another structure for the Aβ dimer was the in-register antiparallel β-sheet conformation (Fig. 1 B) generated by using the PROFASI software.

The selected structures of Sup35 peptide are shown in Fig. 1, C and D. These two dimer structures were revealed by the REMD simulation (see Supporting Material). The monomers in the dimer are oriented in an antiparallel fashion with out-of-register or in-register arrangements (Fig. 1, C and D, respectively). Five backbone H-bonds from the Asn and Gln residues participate in the formation of the out-of-register dimer in Fig. 1 C, and seven backbone H-bonds form the in-register dimer structure (Fig. 1 D). Two Cα atoms from the Cys residues, indicated with the balls, were chosen as the pulling force application points.

MCP of Aβ dimers

A representative MCP force curve obtained for the Aβ dimer in the out-of-register conformation is shown in Fig. 3 A. The structure of the dimer before the rupture is shown above the force curve. The simulations were carried out at a pulling rate of 500 nm/s, which is close to the experimental pulling rate value. Fig. 3 A shows that the dimer undergoes a sharp transition with a rupture force value of 60 pN. Similar simulations were performed for 586 pulling events, and the distribution of the rupture forces is shown in Fig. 3 B. The force distribution is asymmetrical (skewed to the right) with a geometric mean value of 46 ± 1 pN (the geometric mean ± the standard error of geometric mean, as described previously (32)). Similar simulations for the in-register conformation of the Aβ dimer produced larger forces, as shown in Fig. 3 C. The histogram built for the set of 397 simulation runs is shown in Fig. 3 D. The geometric mean value of 178 ± 3 pN is 4-fold greater than the value obtained for the out-of-register conformation. The experimental value of 53 ± 2 pN obtained at the same pulling rate (14) is considerably closer to the computational data for the out-of-register model (46 ± 1 pN). The difference between the experimental results and the simulated value for the out-of-register model is only ∼10% and can be explained by a number of minor factors, such as the exact ionic conditions and experimental errors in the force calibration. In our previous analysis of the rupture of out-of-register Aβ dimers using the SMD approach, we obtained a >10-fold higher rupture value (14). This is because we used a pulling rate of 5 nm/ns, which is 10⁷ times greater than the pulling rates used in MCP simulations and the experiment.

Figure 3.

Rupture force curves and distributions of Aβ peptide at 300 K. (A) Typical force curve for the rupture of the out-of-register dimer. The snapshot of the dimer structure before rupture is above the force curve. (B) Rupture force distribution for the force-induced dissociation of the out-of-register dimer. (C) Typical force curve for the dissociation of the in-register dimer. The snapshot of the dimer structure before the rupture is above the force curve. In A and C, the distance on the x axis of the force plots corresponds to the distances between the Cα atoms of the N-terminal Cys residues. The arrows indicate β-strands, the tubes are random coils, and the dotted lines are H-bonds. (D) Rupture force distribution for the in-register dimer. The force distribution histograms are approximated with PDFs. The bimodal approximation fits the histograms. Individual PDF distributions shown by dotted lines essentially coincide with the overall distributions shown by solid lines. See Fig. S2, in which individual PDF fits are shown in color. To see this figure in color, go online.

The MCP simulations revealed three classes of out-of-register Aβ dimers, which differ in their rupture processes (Fig. 4). The parameters used for the characterization of different structures are the rupture force values, the number of H-bonds, and the β-sheet content. Class I structures are defined by a rupture force value > 20 pN, number of H-bonds ≥ 0, and β-sheet content = 0. Class II structures have a rupture force value > 20 pN, number of H-bonds = 1–5, and β-sheet content > 0. Class III is similar to class II, but the number of H-bonds is >5. In class I dimers (Fig. 4 A), the β structure dissociates before approaching the maximum rupture force, leading to the formation of a non-β structure stabilized by aromatic-aromatic (Ar-Ar) interactions of four Phe residues. This structure is schematically shown in Fig. 4 A. Its dissociation, averaged over 386 events, produces a mean rupture force of 32 ± 1 pN (Fig. 3 B, first peak). Class I dimers are the most representative population of the rupture events in the MCP simulations. In class II (Fig. 4 B), the out-of-register structure retains a few H-bonds before reaching the maximum force, and Ar-Ar interactions contribute to the structural stability of the dimer. In class III dimers (Fig. 4 C), the out-of-register structure undergoes a conformational transition to structures containing a high β-sheet content, with some conformations forming in-register dimers. The conformational transitions occur due to a relatively low pulling rate (500 nm/s) and the fast rate of β-sheet formation, which is on the microsecond timescale (33). These conformational transitions can occur during pulling of the in-register dimers. This model explains the broad distribution for the in-register dimer pulling results (Fig. 3 D). Due to the fast rate of the conformational transitions, dimers with fewer numbers of H-bonds are formed and dissociations occur at low forces. The forces from class II and III structures contribute to the asymmetry of the overall force distributions. Although there is a significant difference between the simulation of out-of-register dimers and experimental conditions (p < 0.01), class I and II rupture conformations correspond to the rupture force values that are close to experimental rupture force values (see distributions in Fig. S2). This finding suggests that a combination of these two types of structures is probed with the experiment.

Figure 4.

Modeling of the rupture process for three classes of structures for the out-of-register Aβ dimer. The simulation was performed at 300 K. (A) The class I structure is characterized by the transient formation of dimers stabilized by aromatic-aromatic (Ar-Ar) interactions. (B) The class II structure contains H-bonding and Ar-Ar interactions. (C) The class III structure is rearranged from the initial structure to form more H-bonds. The arrows indicate β-strands, the tubes are random coils, the stick structures represent Phe residues, and the dotted lines are H-bonds. To see this figure in color, go online.

MCP of Sup35 dimers

Next, we used the MCP simulation to characterize the rupture of Sup35 dimers. The REMD analysis generated two structures (Fig. 1, C and D), and the typical rupture profiles simulated with the MCP approach and experimental method are shown in Fig. 5. The distribution for the rupture forces for the out-of-register Sup35 dimer shown in Fig. 5 A results in a rupture force of 33 ± 1 pN (n = 318). The rupture profile for the out-of-register Sup35 dimer data shows that there is tremendous fluctuation in the dimer structure (Fig. 5 A, inset). Although five H-bonds remain stable (Fig. S3 A), the two extra dangling H-bonds between Tyr of one monomer and Gly of the other monomer are unstable before the rupture event. According to Fig. S3 B, which shows the residue position fluctuations, the terminal Cys residues are floppy and characterized by a relatively large root mean-square fluctuation (RMSF) value (>0.3 nm).

Figure 5.

Rupture force distributions for Sup35 dimers at 300 K. (A) Theoretical data for the dissociation of the out-of-register dimer. (B) Theoretical data for the dissociation of the in-register dimer. (C) Experimental results for the rupture force distribution for Sup35 dimer at a pulling rate of 300 nm/s. Note that the distance in the force curve shown in C (inset) includes the length of the stretching polymer tether used for peptide immobilization (6). The solid lines indicate the overall fit approximation with the bimodal PDF and the dotted lines represent individual PDF fits. The insets are the representative force curves. The models of structures are shown above the force curves in the insets. The arrows indicate β-strands, the tubes are the random coils, and the dotted lines are H-bonds. The distance in the graphs corresponds to the distances between the Cα atoms of the N-terminal Cys residues (black lines in the insets of A and B). To see this figure in color, go online.

A similar analysis was performed for the in-register dimer (Fig. S3, C and D). A representative force curve is shown in Fig. 5 B (inset). Based on 991 rupture events (the distribution is shown in Fig. 5 B), the rupture force was determined to be 58 ± 1 pN. The in-register dimer has seven H-bonds that remain intact until the end of the rupture process. Furthermore, unlike the out-of-register dimers, the number of H-bonds in the in-register dimers remains constant before the rupture starts (Fig. S3 C), and there are no dangling H-bonds. Additionally, the RMSF value of the residues is relatively low (<0.3 nm; Fig. S3 D), suggesting that the swing residues and the dangling H-bonds in the out-of-register dimer are responsible for its reduced conformational stability. Similar to the results for the Aβ dimer structures, the out-of-register dimer with a low β-sheet content constitutes the most representative species, suggesting that these structures were probed in the majority of the force-probing experiments.

The force distribution of the experimental data assembled in Fig. 5 C has a peak value of 42 ± 2 pN that is significantly less than the value obtained for the simulation of the in-register dimer (58 ± 1 pN; p < 0.01), and closer to that obtained for the simulation for the out-of-register dimer (33 ± 1 pN). The comparison between these values is summarized by the bar histogram in Fig. S4. Next, we fitted the force distributions in Fig. 5 with bimodal PDFs. Such an approximation shows that the experimental data (main peak at 33 pN, shoulder peak at 55 pN) correlate well with the theoretical data for the out-of-register model (main peak at 32 pN, shoulder peak at 50 pN). At the same time, the in-register model has a minor first peak at 40 pN and the major second peak at 92 pN. These comparisons are summarized in Table S2 and Fig. S4. This comparative analysis suggests that the out-of-register Sup35 dimer is the predominant structure probed by the SMFS experiment.

The elevated dynamics we observed for Sup35 dimers is also in agreement with the computational analyses described by Srivastava and Balaji (34), who analyzed the dynamics of the Sup35 crystallographic hexamer structure (35). These simulations showed that the crystallographic hexamer is not stable and dissociates in the course of the simulation process, suggesting that additional interactions within a large ensemble of the peptide units are responsible for ensemble stabilization.

Temperature dependence of the stabilities of Sup35 and Aβ dimers

The increased mobility of the terminal residues identified in the comparative structural studies of the dimers under pulling stress indicates differences in the dimers’ stabilities. To evaluate the thermodynamic stabilities of both types of Sup35 and Aβ dimers, we performed MCP simulations for the four structures at temperatures of 288 K and 266 K. The force histograms for Aβ and Sup35 peptides are shown in Figs. S5 and S6, respectively. There is a trend toward higher forces as the temperature decreases, suggesting that thermal fluctuations destabilize the dimers in the force probing. This assumption was confirmed by an analysis in which the fractions of dissociated dimers for both peptides (adopted in-register and out-of-register conformations) were determined from the ratio of the number of simulations with nonrupture events to the total number of simulations. The results are shown in Table S1 and plotted in Fig. 6. They demonstrate that the dissociation fraction increases with temperature, but the association varies depending on the type of peptide and its conformation. The dependence on temperature is less steep for in-register conformations than for out-of-register conformations, and the in-register Aβ dimer is not dependent on the temperatures used in this analysis (Fig. 6 A). This suggests that the in-register structure is stable with respect to the out-of-register structure. The temperature dependence for the out-of-register Sup35 peptide is the steepest (Fig. 6 B), suggesting that the dimer in this conformation is very dynamic.

Figure 6.

Temperature dependence of the fraction of dissociated dimers for the Aβ dimer (A) and the Sup35 dimer (B). The gray dashed lines represent the out-of-register dimers, and the black solid lines represent the in-register dimers.

The higher stability of the out-of-register Aβ dimer compared with the out-of-register Sup35 dimer can be explained by the elevated hydrophobicity of the Aβ peptide and the hydrophilic feature of the Sup35 peptide. Additionally, Aβ contains three charged residues (Lys-16, Glu-22, and Asp-23) that interact within the dimer to contribute to dimer stability. Our previous analysis (14) identified the formation of salt bridges and aromatic interactions as additional stability factors for Aβ dimers.

Dynamics of Sup35 and Aβ dimers and the aggregation process

The computational analysis for Sup35 and Aβ peptides showed that the dimers are capable of forming out-of-register and in-register arrangements. However, the comparison with the experimental data led to the conclusion that both peptides in the AFM probing experiments assemble as dimers in an out-of-register alignment. Given the higher stability of the in-register dimer structure compared with the out-of-register structure, one would expect that the formation with the most stable structure would occur in AFM probing experiments. The dimers could undergo the transition into the in-register conformation before they grow into larger oligomers. This is supported by the observation of rupture events with forces considerably exceeding that found for the out-of-register conformation, although the yield of these events is very low, in the range of a percent (6, 14).

According to the energy landscape 2D diagram for Sup35 produced by the REMD simulation (Fig. S1), there are two major local energy minima corresponding to the most stable configurations. Therefore, it is reasonable to assume that out-of-register structures are kinetically trapped, and the dimer can adopt the most stable conformation over time after passing a barrier between the two energy minima. In our previous MD simulations of Aβ peptide, we observed the formation of an in-register dimer configuration that began with the out-of-register conformation (14). This transition required full dissociation of the dimer followed by rearrangement of the peptide chains, enabling the in-register antiparallel orientation. This was observed in an extended MD simulation process (up to ∼2 μs), confirming the kinetic trap of the out-of-register conformations. Therefore, the kinetically trapped out-of-register conformation can self-assemble and form higher-order oligomers without changing the out-of-register conformation. Oligomers assembled with the in-register dimer should be structurally different. However, we speculate that the first types of oligomers can undergo structural transitions that form the second type of oligomers in the in-register conformation. This is supported by the recent observation of out-of-register conformations of β2-microglobulin hexapeptide in crystals (36). The model of conformational transitions within oligomers was proposed in a study of β-lactoglobulin aggregation (37).

Since the analysis described above focused on antiparallel dimers, we thought it would be interesting to compare this analysis with one for the parallel arrangement of monomers. We modeled the parallel arrangement of Aβ peptide and performed rupture simulations. A representative force curve is shown in Fig. S7. There are peaks at ∼2 nm, ∼4 nm, and ∼6 nm corresponding to the stepwise dimer unzipping, but their amplitudes are slightly above the noise level. These data are very different from those obtained by pulling of antiparallel dimers with well-defined peaks (Figs. 3 and 5).

Conclusions

Overall, our simulations revealed that the dimers formed by Sup35 and Aβ have structural variability and differences in their dynamics. However, we also identified some similarities in their structures. The formation of dimers with diverse structures can lead to different aggregation pathways and produce oligomers with different structures that may have physiological significance. Although we used the MCP approach to analyze short peptides, it may be possible to apply this approach to larger systems, as demonstrated by the analysis of titin and ubiquitin proteins. The development of the modified MCP computational approach may facilitate the structural characterization of large protein systems probed by AFM.

Editor: Elizabeth Rhoades.

Supporting Citations

Refs (38, 39, 40, 41) appear in the Supporting Material.